Share your thoughts, 1 month free Claude Pro on usSee more

Web-based Reasoning on WebShop

34.8Average Reasoning Length (tokens)

Teacher (LLaMA-13B)

Updated 1mo ago

Evaluation Results

Method	Links
Teacher (LLaMA-13B) 2025.05		34.8
Ours (LLaMA-7B) 2025.05		34.9
Ours (OPT-6.7B) 2025.05		35.1
Teacher (OPT-13B) 2025.05		35.9
Ours (OPT-2.7B) 2025.05		36.2
Token-level-OPT-6.7B 2025.05		37.2
Token-level-7B 2025.05		37.5
SeqKD (7B) 2025.05		37.8
Ours (OPT-1.3B) 2025.05		38
SeqKD (6.7B) 2025.05		38.2
KD (7B) 2025.05		38.9
Token-level-2.7B 2025.05		39
KD (6.7B) 2025.05		39.1
SeqKD (2.7B) 2025.05		39.8
KD (2.7B) 2025.05		40.7
Token-level-1.3B 2025.05		41.8
SeqKD (1.3B) 2025.05		42.9
KD (1.3B) 2025.05		44.1